Israel is financing a $46.5 million operation to influence how artificial intelligence chatbots such as ChatGPT and Claude talk about the war in Gaza, reports Drop Site News on Tuesday.
Spearheaded by former Trump campaign manager Brad Parscale through his digital PR firm Clock Tower X, hundreds of articles have been published since October 2025 on a network of seemingly independent websites. Disinformation researchers say these sites are primarily for search engines, web crawlers, and the data sets used to train large language models.
Other outlets such as the Quincy Institute and Middle East Monitor confirm the attempt, raising new questions about ways state actors attempt to manipulate AI systems that millions of people now consult for information.
A Network Designed for Machines, Not Readers:
Investigators found that Clock Tower X’s websites attract only a few hundred human visitors each month despite costing millions to operate. Instead, researchers say the network primarily targets AI crawlers that ingest content for training large language models. Published articles falsely claim that most journalists in Gaza were linked to terrorist groups.
One site promotes US-Israeli technology collaboration, while another claims footage from Gaza is staged and denies reports of famine. Analysts found that 10 of the network’s sites were crawled 912 times by Common Crawl between January and June 2026, with the frequency increasing sharply as the campaign expanded, indicating they were optimized for AI training datasets rather than human readers.
Is It Working?
Disinformation researchers say the strategy seems to be working. Both were able to retrieve and cite the sites during live web searches, but testing did not turn up direct evidence that either of the two systems was trained on the network’s material. That distinction is important. Shaping the search results of a chatbot is not the same as permanently implanting content into the training data of a model, but experts warn both paths can influence what users ultimately view as fact.
Hervé Letoqueux, head of the disinformation monitoring group Check, said: “First, they warned that although researchers can confirm that some AI models have been exposed to the network’s content, it is still difficult to measure how much this has influenced the responses of individual chatbots. Common Crawl’s centrality to training artificial intelligence systems makes it an appealing target for actors looking to shape outcomes, he said.
The Strategy of Content Marketing
Unlike conventional propaganda efforts that use emotional or incendiary rhetoric, the Clock Tower X network reportedly prefers a measured, ”factual” tone that is more AI citeable.
“Content that’s structured and sourced in a certain way is more likely to be retrieved when somebody’s searching on an AI platform”, Scott Stouffer, chief technology officer of Market Brew, a firm that influences language models and in which Parscale has invested. The point, he suggested, was less to win arguments than to make sure the pro-Israel frame was just there when the relevant questions were asked.
Researchers also point out that many articles on the network include measured caveats or nuance, a technique that could help the content blend in with legitimate journalism and not raise obvious flags as propaganda.
Broader Implications for AI and Information Warfare
The findings arrive as governments and political operatives are under greater scrutiny for how they are adapting influence campaigns to an information landscape increasingly driven by AI. This operation does not target social media feeds or television audiences but the invisible infrastructure crawlers, indexes and training pipelines that increasingly determine what artificial intelligence systems tell users. Other content in the network asserts that Israel follows international law and permits humanitarian aid, positions that researchers say contradict extensive documentation from United Nations agencies and international rights groups about the conflict.
For now, AI companies have not publicly detailed specific countermeasures against this kind of targeted content seeding. As chatbots became a primary source of information for many users, the episode highlights a growing vulnerability: the same openness that makes data sets like Common Crawl useful for building AI also makes them susceptible to coordinated manipulation by well-funded actors.
Related Stories:
AI Data Centers Consume More Power Than Most Nations: UN Report
Instagram users face shadow ban over Pro-Palestinian content
